How to make a DLL in principle - MQL4 and MetaTrader 4

liza 2008.06.18 09:27 #51

YuraZ писал (а) >>

Another option but on MQL

just trying to work on M1 shows the current trend direction

exit on the 3rd, in theory you can not teach at startup - but immediately take ready Weights - teach in advance

---

9 inputs to which the normalized difference of muwings is fed

---

out-1 out-2 out-3 output

0.9 0.01 0.01 trend up

0.01 0.9 0.01 flet

0.01 0.01 0.9 trend Down

---

I plan to add 3-4 timeframes and calculate them all at once!

PNN work very fast when weights are cranked up

---

Experimental test code - please don't kick me!

I have tried to run YZ_BETTER_HC_1_1.mq4 on different TFs with default parameters, not a single trade. Maybe I did something wrong?

Yuriy Zaytsev 2008.06.18 11:52 #52

liza писал (а) >>

Could not find a dll file. If you can, please tell me how to create it.

it's not a DLL, it's a VC++ 6 project

You can make a DLL in principle!

You need to see an example of how to make a DLL.

example

#define WIN32_LEAN_AND_MEAN 
#define MT4_EXPFUNC __declspec(dllexport)
#include <windows.h>
 
BOOL WINAPI DllMain(HANDLE hModule,DWORD ul_reason_for_call,LPVOID lpReserved)
{
    switch(ul_reason_for_call)
    {
        case DLL_PROCESS_ATTACH:
        case DLL_THREAD_ATTACH:
        case DLL_THREAD_DETACH:
        case DLL_PROCESS_DETACH:
        break;
    }
    return(TRUE);
}
// тут наверно что то хотите в свою сеть передавать
//
MT4_EXPFUNC int __stdcall LIZA_PNN(  int x1, int x2,double x3, double *_inp,   double *_out )
{
//
// тут код Вашей сети
// и на вход вы ей что то передали из MT4
 
//
// видимо вернуть надо в MT что то
//
   _out[0] = ???;
   _out[1] = ???;
   _out[3] = ???;
 
    return(0);
}

DLL creation example

http://ai.xss.ru/progs/optim/DLLOpisanie.htm

Only the METAQUOTES requirement must be met when you create a DLL.

Create your own MetaTrader Socket programming in MT4. Machine learning in trading:

Yuriy Zaytsev 2008.06.18 11:54 #53

liza писал (а) >>

it's just an attempt to input a muving difference

and get the output trying to hold the input direction !

i.e. the net will just say SIGNAL UP or SIGNAL DOWN!

it will not give you an entry point! it will help you hold the trade

entrances can be done differently!

Machine learning in trading: [WARNING CLOSED!] Any newbie Not looking for investors

Yuriy Zaytsev 2008.06.18 12:07 #54

FION писал (а) >>
Maybe to go the way of connecting DLL in C++ to speed up MQL. Write some universal templates-functions in C++, though it's kind of the same as NSDT. By the way, BETTER wrote everything in C++ and translated the final result into MQL.

The NS linkage has in principle already been solved... by many ... with success and a long time ago

just an attempt to code NEUROSET with your own hands with an opportunity to "play around" with the network scheme - structure, learning algorithm, etc...

to create something better, more manageable, and thus more predictable and easily modifiable than black boxes like NS, SOLUTION

and there is no way such bundles get into the competition ... And the authors of such systems do not always understand how a network works

for they are black boxes...

DLL or external module in C++ - just for the sake of speed - with integration of their network with MT4

it makes no sense to translate the NS into MQL4, except perhaps for the autumn and annual competitions ...

---

Using Neural Networks in Neural network NS + indicators. Experiment.

Andrey 2008.06.18 15:31 #55

YuraZ писал (а) >>

The NS linkage has in principle already been solved... by many ... with success and for a long time ...

just an attempt to code the NEUROSET with your own hands ... with the ability to "play around" with the network scheme - structure, learning algorithm, etc...

to create something better, more manageable, and thus more predictable and easily modifiable than black boxes like NSH, SOLUTION

and there is no way such a bunch will get to the competition... And the authors of such systems do not always understand well how the net works

for they see it as a black box...

DLL or an external module in C++ - only because of the speed - with the integration of their network with MT4

There is no sense to translate NS to MQL4 - well, maybe for the autumn and probably annual competitions ...

--

There is a software product NeuroSolution that allows implementing networks of any architecture, it is not a black box like NS, if you think so, I solved this problem with the help of a unified dll, using this module we can connect networks created in any architecture to MetaTrader http://www.myfolder.nm.ru/nsdt_and_ns.htm and there is no need to program and debug the network in mql, besides there are great possibilities to build and configure the structure of networks, in this case the network training takes place not in the Metatrader, you can even run the game in mql.

It is impossible to transfer it to mql as well, or at least it will be technically very difficult, but for the work and searching the optimal configuration it is the best option, in my opinion.

What is a black box for you, if weights and program code, or how the search of weights occurs during backtracking, I don't think it's so important to understand how grids work, to know what value is multiplied by what and where the result comes to after next addition... i don't think it could be interesting in terms of understanding, certain structure of network, training method impose corresponding functional features that have a qualitative impact on the result and speed, which are handled in the process of design, number of neurons in the layer, speed and acceleration in this case can be optimized via genetic algorithm, which is also available in a soluto

Another Algorithms to optimization/Trainning Probabilistic neural networks, packages How to create Neural

Artem Titarenko 2008.06.18 16:33 #56

I completely agree with Garfish on the issue of neural network development. NeuroSolution is really not a black box compared to NeuroShellDT...

In my opinion, NeuroSolution is almost ideal software for creating NS. Just to use it, unlike NSDT, you need pretty good neural network knowledge...

And you can hardly beat NeuroSolution with MQL4...

THE IDEA EXCHANGE Neuromongers, don't pass by Probabilistic neural networks, packages

liza 2008.06.19 18:18 #57

liza писал (а) >>

YuraZ wrote (a) >>

it's not a DLL, it's a VC++ 6 project

You can make a DLL in principle!

You need to see an example of how to make a DLL.

example

DLL creation example

http://ai.xss.ru/progs/optim/DLLOpisanie.htm

Only the METAQUOTES requirement must be met when you create a DLL.

liza wrote (a) >>

I did everything as described. The .exe, .lib etc. were created but the .dll still didn't appear. Could you tell me what's the problem?

Andrey 2008.06.19 18:50 #58

liza писал (а) >>

I did everything as described. The .exe, .lib, etc. were created, but the .dll still didn't appear. Can you tell me what the problem is?

The problem is you don't understand what you're doing.

Yuriy Zaytsev 2008.06.20 06:44 #59

liza писал (а) >>

I did everything as described. The .exe, .lib, etc. are created, but the .dll is still not there. Could you tell me what's wrong?

DLL creation example

http://ai.xss.ru/progs/optim/DLLOpisanie.htm

liza 2008.06.20 07:17 #60

YuraZ писал (а) >>

DLL creation example

http://ai.xss.ru/progs/optim/DLLOpisanie.htm

This is the example I tried to make. No errors, when "Build All" creates everything except the .dll.

Neural network in the form of a script - page 6